@stll/folio-core 0.45.0 → 0.46.0

package/dist/prosemirror/conversion/fromProseDoc.js

@@ -1,4 +1,7 @@
+import { joinCommentRangesAcrossParagraphs } from "../../docx/commentRangeJoin.js";
+import { completeCommentReferences } from "../../docx/commentReferenceCompletion.js";
 import { EDITED_PREVIEW_FINGERPRINT, imageRawXmlFingerprint } from "../../docx/imageRawXml.js";
+import { isInlineSdtContent } from "../../docx/inlineWrapperContent.js";
 import { isNumberingReference } from "../../docx/numberingReference.js";
 import { PROSE_PARAGRAPH_SOURCE_CONTRACT_ATTR, ParagraphPropertySourceValidationError, copyDocumentParagraphPropertySourceContract, copyDocumentParagraphPropertySources, copyParagraphPropertyCapture, copyParagraphPropertySource, decodeTableCellParagraphSourcePayload, getDocumentParagraphPropertySourceContract, getParagraphPropertySource, getParagraphPropertySourceCandidate, getParagraphPropertySourceToken, getParagraphPropertySourceTransferId, getProseDocumentParagraphPropertySourceContract, getProseParagraphPropertySourceToken, isParagraphPropertySourceToken, linkParagraphPropertySourceCandidate, paragraphPropertySourceBelongsToDocument, paragraphPropertySourceTokenMatchesContract, recreateProseNodeWithParagraphPropertySource, restoreTableCellsWithParagraphPropertySources, visitDocumentStoryParagraphs, visitTableCellParagraphPropertySourceBindings } from "../../docx/paragraphPropertySource.js";
 import { visitDocxParagraphs } from "../../docx/paragraphTraversal.js";
@@ -11,12 +14,16 @@ import { canonicalJson } from "../../utils/canonicalJson.js";
 import { normalizeHorizontalScalePercent } from "../../utils/horizontalScale.js";
 import { mergeTextFormatting } from "../../utils/textFormattingMerge.js";
 import { emuToPixels, emuToStrokePixels, pixelsToEmu } from "../../utils/units.js";
-import { expectBlockSdtAttrs, expectCharacterSpacingMarkAttrs, expectCharacterStyleMarkAttrs, expectCommentMarkAttrs, expectEmphasisMarkAttrs, expectFieldAttrs, expectFontFamilyMarkAttrs, expectFontSizeMarkAttrs, expectFootnoteRefMarkAttrs, expectHardBreakAttrs, expectHighlightMarkAttrs, expectHyperlinkMarkAttrs, expectImageAttrs, expectLanguageMarkAttrs, expectMathAttrs, expectPageBreakRunAttrs, expectPageBreakRunOwnerMarkAttrs, expectParagraphAttrs, expectRunFormattingOverrideMarkAttrs, expectRunPropertyChangeMarkAttrs, expectRunShadingMarkAttrs, expectSdtAttrs, expectShapeAttrs, expectStrikeMarkAttrs, expectSymbolAttrs, expectTabAttrs, expectTableAttrs, expectTableCellAttrs, expectTableRowAttrs, expectTextBoxAttrs, expectTextColorMarkAttrs, expectTextEffectMarkAttrs, expectTrackedChangeMarkAttrs, expectUnderlineMarkAttrs } from "../attrs/index.js";
+import { expectBlockSdtAttrs, expectCharacterSpacingMarkAttrs, expectCharacterStyleMarkAttrs, expectCommentMarkAttrs, expectEmphasisMarkAttrs, expectFieldAttrs, expectFontFamilyMarkAttrs, expectFontSizeMarkAttrs, expectFootnoteRefMarkAttrs, expectHardBreakAttrs, expectHighlightMarkAttrs, expectHyperlinkMarkAttrs, expectImageAttrs, expectInlineWrapperMarkAttrs, expectLanguageMarkAttrs, expectMathAttrs, expectPageBreakRunAttrs, expectPageBreakRunOwnerMarkAttrs, expectParagraphAttrs, expectPreservedBlockAttrs, expectPreservedXmlAttrs, expectRunFormattingOverrideMarkAttrs, expectRunPropertyChangeMarkAttrs, expectRunShadingMarkAttrs, expectSdtAttrs, expectShapeAttrs, expectStrikeMarkAttrs, expectSymbolAttrs, expectTabAttrs, expectTableAttrs, expectTableCellAttrs, expectTableRowAttrs, expectTextBoxAttrs, expectTextColorMarkAttrs, expectTextEffectMarkAttrs, expectTrackedChangeMarkAttrs, expectUnderlineMarkAttrs } from "../attrs/index.js";
+import { readAuthoredTransform } from "../authoredTransformAttrs.js";
 import { autospacingMatchesBase, hasAutospacingBaseSide } from "../autospacingBase.js";
 import { bookmarkBoundaryDisplacement, expectBookmarkBoundaryAttrs } from "../bookmarkBoundaryAttrs.js";
 import { paragraphRejectAttrPatch, paragraphRejectOriginalFormatting, removeParagraphPropertyChanges } from "../commands/propertyChangeScope.js";
+import { expectCommentReferenceAttrs } from "../commentReferenceAttrs.js";
+import { INLINE_WRAPPER_MARK_NAME } from "../extensions/marks/InlineWrapperExtension.js";
 import { applyRunFormattingOverrideAttrs, buildRunFormattingOverrideAttrs } from "../extensions/marks/RunFormattingOverrideExtension.js";
 import { textFormattingToMarks } from "../extensions/marks/markUtils.js";
+import { inlineWrapperMember, inlineWrapperStackKey } from "../inlineWrapperStack.js";
 import { listRenderingFromAttrs as listRenderingFromAttrs$1 } from "../listRenderingAttrs.js";
 import { resolveNumberedRefFields } from "../numberedRefFields.js";
 import { directParagraphAlignment } from "../paragraphAlignment.js";
@@ -26,6 +33,7 @@ import { RUN_FORMATTING_MARK_NAMES } from "../runFormattingMarkNames.js";
 import { authoredRunFormattingFromAttrs, hasAuthoredRunFormattingProvenance } from "../runFormattingProvenance.js";
 import { paragraphFormattingForRun, paragraphRunStyleContext, resolveEffectiveRunStyleFormatting, suppressParagraphMarkFormatting } from "../runStyleFormatting.js";
 import { schema } from "../schema/index.js";
+import { PRESERVED_XML_LEVELS } from "../schema/nodes.js";
 import { parseShapeGeometryAdjustments } from "../shapeGeometryAdjustments.js";
 import { expectTextBoxAnchorAttrs } from "../textBoxAnchorAttrs.js";
 import { assertValidProseMirrorDocument } from "../validation.js";
@@ -51,19 +59,6 @@ function normalizeShapeOutlineStyle(style) {
 	if (style in OUTLINE_STYLE_CSS_ALIASES) return OUTLINE_STYLE_CSS_ALIASES[style];
 	return narrowEnum(style, ShapeOutlineStyleSchema);
 }
-function parseTransformAttr(transformStr) {
-	if (!transformStr) return;
-	const transform = {};
-	const rotateMatch = /rotate\((?<deg>[-\d.]+)deg\)/u.exec(transformStr);
-	if (rotateMatch) {
-		const rotation = Number.parseFloat(rotateMatch.groups["deg"]);
-		if (Number.isFinite(rotation)) transform.rotation = rotation;
-	}
-	if (transformStr.includes("scaleX(-1)")) transform.flipH = true;
-	if (transformStr.includes("scaleY(-1)")) transform.flipV = true;
-	if (transform.rotation === void 0 && !transform.flipH && !transform.flipV) return;
-	return transform;
-}
 function imagePositionFromAttrs(attrs) {
 	const horizontalPosition = attrs?.horizontal;
 	const verticalPosition = attrs?.vertical;
@@ -236,6 +231,8 @@ function fromProseDoc(pmDoc, baseDocument) {
 	if (proseContractAttribute !== null && proseContractAttribute !== void 0 && proseContract === null || (baseContract ?? null) !== proseContract) throw sourceValidationError("contract_mismatch", "The ProseMirror document does not match its paragraph-property source document.");
 	const tokenSources = baseContract && proseContract && baseDocument ? validateParagraphPropertySourceTokens(pmDoc, baseDocument, baseContract) : null;
 	const blocks = extractBlocks(pmDoc, "resolve", baseDocument?.package.styles ? createStyleEngine(baseDocument.package.styles) : null);
+	joinCommentRangesAcrossParagraphs(blocks);
+	completeCommentReferences(blocks);
 	const linkedSources = restoreLinkedParagraphPropertySources(blocks);
 	if (tokenSources) restoreParagraphPropertySourcesByToken(blocks, tokenSources);
 	else if (baseDocument) restoreParagraphPropertySources(blocks, baseDocument.package.document.content, linkedSources.targets, linkedSources.sources);
@@ -452,42 +449,9 @@ function materializeNumberedRefValues(doc) {
 	};
 	return visit(doc);
 }
-const SERIALIZER_EMPTY_PAGE_FIELD_RESULT = "1";
-const fieldResultHasExplicitPageBreak = (node) => extractParagraphContent(node, void 0, void 0, /* @__PURE__ */ new Map(), false).filter((content) => content.type === "run" || content.type === "hyperlink").some((content) => {
-	return (content.type === "run" ? [content] : content.children).some((run) => run.type === "run" && run.content.some((runContent) => runContent.type === "break" && runContent.breakType === "page"));
-});
-/**
-* PAGE and NUMPAGES have a stable visible fallback. Other empty fields may
-* intentionally have no result (for example an empty TOC); inventing a space
-* changes their authored result on every save/reopen cycle. A field whose
-* result already carries an explicit page break owns that break's run, so the
-* fallback would displace it.
-*/
-function materializeSerializerFieldFallbacks(doc) {
-	const visit = (node) => {
-		if (node.type.name === "field" || node.type.name === "structuredField") {
-			const attrs = expectFieldAttrs(node);
-			return (attrs.fieldType === "PAGE" || attrs.fieldType === "NUMPAGES") && !attrs.displayText && !fieldResultHasExplicitPageBreak(node) ? recreateProseNodeWithParagraphPropertySource(node, { attrs: {
-				...node.attrs,
-				displayText: SERIALIZER_EMPTY_PAGE_FIELD_RESULT
-			} }) : node;
-		}
-		if (node.childCount === 0) return node;
-		const children = [];
-		let changed = false;
-		node.forEach((child) => {
-			const mappedChild = visit(child);
-			children.push(mappedChild);
-			changed ||= mappedChild !== child;
-		});
-		return changed ? recreateProseNodeWithParagraphPropertySource(node, { content: Fragment.fromArray(children) }) : node;
-	};
-	return visit(doc);
-}
-function extractBlocks(inputDoc, refResolution = "resolve", styleResolver = null, emptyFieldResult = "serializerFallback") {
+function extractBlocks(inputDoc, refResolution = "resolve", styleResolver = null) {
 	const strippedDoc = stripSuggestedProvenance(inputDoc, styleResolver);
-	const refResolvedDoc = refResolution === "resolve" ? materializeNumberedRefValues(strippedDoc) : strippedDoc;
-	const pmDoc = emptyFieldResult === "serializerFallback" ? materializeSerializerFieldFallbacks(refResolvedDoc) : refResolvedDoc;
+	const pmDoc = refResolution === "resolve" ? materializeNumberedRefValues(strippedDoc) : strippedDoc;
 	const blocks = [];
 	const textBoxAnchorMarkers = /* @__PURE__ */ new Map();
 	const documentCounts = buildDocumentTrackedChangeCounts(pmDoc);
@@ -530,15 +494,69 @@ function extractBlocks(inputDoc, refResolution = "resolve", styleResolver = null
 			pendingPageBreaks = 0;
 		} else if (node.type.name === "blockSdt") {
 			if (pendingPageBreaks > 0 && !appendPendingPageBreaksToPreviousParagraph()) flushPendingPageBreaks();
-			blocks.push(convertPMBlockSdt(node, styleResolver, emptyFieldResult));
+			blocks.push(convertPMBlockSdt(node, styleResolver));
+			previousStandaloneTextBox = null;
+		} else if (node.type.name === "preservedBlock") {
+			blocks.push({
+				type: "preservedBlock",
+				xml: expectPreservedBlockAttrs(node).xml
+			});
 			previousStandaloneTextBox = null;
 		}
 	});
 	if (pendingPageBreaks > 0 && !appendPendingPageBreaksToPreviousParagraph()) flushPendingPageBreaks();
 	removeUnresolvedTextBoxAnchors(blocks, textBoxAnchorMarkers);
+	keepOneAttributeRemainderPerRecord(blocks);
 	return blocks;
 }
-function convertPMBlockSdt(node, styleResolver, emptyFieldResult) {
+/**
+* The attribute remainder follows the record it was authored on, and only it.
+*
+* ProseMirror copies a node's attrs when a command splits it, so pressing
+* Enter in the middle of a paragraph produces two nodes holding the *same*
+* remainder array. Writing it back on both would give the new half a
+* revision-session id nobody assigned to it: `w:rsidR` says which editing
+* session wrote this paragraph, and a copy of it is a claim about history the
+* author did not make. The rule is that the half that comes first in document
+* order keeps the authored identity, and the other half is a new record with
+* no remainder — the same answer a paragraph the editor created from scratch
+* gets, which is none.
+*
+* Reference identity is what tells the two cases apart, and it is exact: two
+* records that each parsed their own attributes hold different arrays however
+* equal their contents, and only a copy made by the editor shares one.
+*/
+const keepOneAttributeRemainderPerRecord = (blocks) => {
+	const seen = /* @__PURE__ */ new WeakSet();
+	const keepFirst = (record) => {
+		const remainder = record.preservedAttributes;
+		if (remainder === void 0) return;
+		if (seen.has(remainder)) {
+			delete record.preservedAttributes;
+			return;
+		}
+		seen.add(remainder);
+	};
+	const walk = (content) => {
+		for (const block of content) switch (block.type) {
+			case "paragraph":
+				keepFirst(block);
+				break;
+			case "table":
+				for (const row of block.rows) {
+					keepFirst(row);
+					for (const cell of row.cells) walk(cell.content);
+				}
+				break;
+			case "blockSdt":
+				walk(block.content);
+				break;
+			default: break;
+		}
+	};
+	walk(blocks);
+};
+function convertPMBlockSdt(node, styleResolver) {
 	const attrs = expectBlockSdtAttrs(node);
 	const properties = { sdtType: attrs.sdtType };
 	if (attrs.alias) properties.alias = attrs.alias;
@@ -556,7 +574,7 @@ function convertPMBlockSdt(node, styleResolver, emptyFieldResult) {
 	if (attrs.rawEndPropertiesXml) properties.rawEndPropertiesXml = attrs.rawEndPropertiesXml;
 	if (attrs.rawSdtChildrenBeforeContent) properties.rawSdtChildrenBeforeContent = attrs.rawSdtChildrenBeforeContent;
 	if (attrs.rawSdtChildrenAfterContent) properties.rawSdtChildrenAfterContent = attrs.rawSdtChildrenAfterContent;
-	const extracted = extractBlocks(node.type.schema.node("doc", null, node.content), "inherit", styleResolver, emptyFieldResult);
+	const extracted = extractBlocks(node.type.schema.node("doc", null, node.content), "inherit", styleResolver);
 	return {
 		type: "blockSdt",
 		properties,
@@ -594,6 +612,7 @@ function appendTextBoxBlock(blocks, node, options) {
 		if (!mergeTextBoxIntoTrailingInlineSdts(options.previousStandaloneTextBox.paragraph, paragraph, attrs._docxInlineSdts ?? [])) options.previousStandaloneTextBox.paragraph.content.push(...paragraph.content);
 		return options.previousStandaloneTextBox;
 	}
+	if (attrs._docxPlacement === "standalone" && attrs._preservedAttributes?.length) paragraph.preservedAttributes = attrs._preservedAttributes;
 	prependPageBreaks(paragraph, options.pendingPageBreaks);
 	blocks.push(paragraph);
 	return attrs._docxPlacement === "standalone" && attrs._docxGroupId ? {
@@ -647,7 +666,7 @@ function findTextBoxShapeRun(content) {
 			if (item.content.some((runContent) => runContent.type === "shape" && runContent.shape.shapeType === "textBox")) return item;
 			continue;
 		}
-		if (item.type === "inlineSdt" || item.type === "hyperlink" || item.type === "insertion" || item.type === "deletion" || item.type === "moveFrom" || item.type === "moveTo") {
+		if (item.type === "inlineSdt" || item.type === "hyperlink" || item.type === "inlineWrapper" || item.type === "insertion" || item.type === "deletion" || item.type === "moveFrom" || item.type === "moveTo") {
 			const run = findTextBoxShapeRun(item.type === "hyperlink" ? item.children : item.content);
 			if (run) return run;
 		}
@@ -663,6 +682,7 @@ function replaceTextBoxAnchorInBlocks(blocks, marker, textBoxRun) {
 			for (const row of block.rows) for (const cell of row.cells) if (replaceTextBoxAnchorInBlocks(cell.content, marker, textBoxRun)) return true;
 			continue;
 		}
+		if (block.type === "preservedBlock") continue;
 		if (replaceTextBoxAnchorInBlocks(block.content, marker, textBoxRun)) return true;
 	}
 	return false;
@@ -685,7 +705,7 @@ const textBoxAnchorHost = (item) => {
 		case "deletion":
 		case "moveFrom":
 		case "moveTo":
-		case "bidiWrapper": return item.content;
+		case "inlineWrapper": return item.content;
 		case "run":
 		case "complexField":
 		case "bookmarkStart":
@@ -697,7 +717,8 @@ const textBoxAnchorHost = (item) => {
 		case "moveFromRangeEnd":
 		case "moveToRangeStart":
 		case "moveToRangeEnd":
-		case "mathEquation": return;
+		case "mathEquation":
+		case "preservedInline": return;
 		default: panic(`Unsupported paragraph content: ${JSON.stringify(item)}`);
 	}
 };
@@ -730,6 +751,7 @@ function removeTextBoxAnchorFromBlocks(blocks, marker) {
 			for (const row of block.rows) for (const cell of row.cells) if (removeTextBoxAnchorFromBlocks(cell.content, marker)) return true;
 			continue;
 		}
+		if (block.type === "preservedBlock") continue;
 		if (removeTextBoxAnchorFromBlocks(block.content, marker)) return true;
 	}
 	return false;
@@ -825,6 +847,7 @@ function convertPMParagraph(node, documentCounts, textBoxAnchorMarkers, styleRes
 	else if (attrs.sectionBreakType) paragraph.sectionProperties = { sectionStart: attrs.sectionBreakType };
 	if (attrs._propertyChanges && attrs._propertyChanges.length > 0) paragraph.propertyChanges = attrs._propertyChanges.map(propertyChangeFromAttrs);
 	if (attrs.pPrMark) paragraph.pPrMark = attrs.pPrMark;
+	if (attrs._preservedAttributes && attrs._preservedAttributes.length > 0) paragraph.preservedAttributes = attrs._preservedAttributes;
 	linkParagraphPropertySourceCandidate(paragraph, node);
 	return paragraph;
 }
@@ -1017,6 +1040,70 @@ function createTrackedRunWrapper(type, info, child) {
 		content
 	};
 }
+/**
+* Comment ids in the order their markers are written at a shared boundary.
+*
+* Two ranges that begin or end at the same position have no order in the
+* editor, where each is a mark, so every emission site orders them by id: a
+* boundary written in the order the marks happened to open is a different
+* document each time the same one is saved.
+*/
+const byCommentId = (ids) => [...ids].toSorted((a, b) => a - b);
+/** The wrappers `node` sits inside, outermost first; empty when it sits in none. */
+const inlineWrapperStackOf = (node) => {
+	const mark = node.marks.find((candidate) => candidate.type.name === INLINE_WRAPPER_MARK_NAME);
+	return mark ? expectInlineWrapperMarkAttrs(mark).stack : [];
+};
+const isRevisionWrapper = (item) => item.type === "insertion" || item.type === "deletion" || item.type === "moveFrom" || item.type === "moveTo";
+/** `stack` closed around `content`, outermost layer first. */
+const nestInlineWrappers = (stack, content) => {
+	const innermost = stack.at(-1);
+	if (innermost === void 0) panic("An inline wrapper group carries no layer");
+	let wrapper = inlineWrapperMember(innermost, content);
+	for (const layer of stack.slice(0, -1).toReversed()) wrapper = inlineWrapperMember(layer, [wrapper]);
+	return wrapper;
+};
+/**
+* The emitted sequence with each wrapper group nested back into the wrappers
+* `toProseDoc` lifted off it.
+*
+* The revision stays outermost. folio already writes a revision outside the
+* hyperlink it spans, the parse leg is revision-owned, and accepting or
+* rejecting one is a range operation over the revision's own content; a
+* wrapper placed outside two revisions would also have to mint one revision id
+* per wrapper. So a revision in the group keeps its place and takes the nest
+* inside it, and the items around it share a nest of their own.
+*
+* A group with nothing in it writes no wrapper: an authored wrapper whose
+* content was deleted or rejected is gone, and an empty one says nothing.
+*/
+const nestInlineWrapperGroups = (items, groups) => {
+	const nested = [];
+	for (const [index, group] of groups.entries()) {
+		const slice = items.slice(group.start, groups[index + 1]?.start ?? items.length);
+		if (group.stack.length === 0) {
+			nested.push(...slice);
+			continue;
+		}
+		let pending = [];
+		const closeNest = () => {
+			if (pending.length === 0) return;
+			nested.push(nestInlineWrappers(group.stack, pending));
+			pending = [];
+		};
+		for (const item of slice) {
+			if (!isRevisionWrapper(item)) {
+				pending.push(item);
+				continue;
+			}
+			closeNest();
+			if (item.content.length > 0) item.content = [nestInlineWrappers(group.stack, item.content)];
+			nested.push(item);
+		}
+		closeNest();
+	}
+	return nested;
+};
 function extractParagraphContent(paragraph, _documentCounts, emptyHyperlinks, textBoxAnchorMarkers, skipLeadingRenderedPageBreak = false, inheritedFormattingOverride) {
 	const content = [];
 	const paragraphStyleContext = paragraph.type.name === "paragraph" ? paragraphRunStyleContext(paragraph) : void 0;
@@ -1033,6 +1120,11 @@ function extractParagraphContent(paragraph, _documentCounts, emptyHyperlinks, te
 	})).toSorted((left, right) => left.attrs.offset - right.attrs.offset || left.order - right.order);
 	let nextEmptyHyperlink = 0;
 	let leadingRenderedPageBreakPending = skipLeadingRenderedPageBreak;
+	const wrapperGroups = [{
+		start: 0,
+		stack: []
+	}];
+	let wrapperGroupKey = "";
 	let currentRun = null;
 	let currentMarksKey = null;
 	let currentHyperlink = null;
@@ -1086,14 +1178,14 @@ function extractParagraphContent(paragraph, _documentCounts, emptyHyperlinks, te
 		if (toClose.length === 0 && toOpen.length === 0) return;
 		flushCurrentInline();
 		currentTrackedChange = void 0;
-		for (const commentId of toClose.toSorted((a, b) => a - b)) {
+		for (const commentId of byCommentId(toClose)) {
 			content.push({
 				type: "commentRangeEnd",
 				id: commentId
 			});
 			openedComments.delete(commentId);
 		}
-		for (const commentId of toOpen.toSorted((a, b) => a - b)) {
+		for (const commentId of byCommentId(toOpen)) {
 			content.push({
 				type: "commentRangeStart",
 				id: commentId
@@ -1111,13 +1203,45 @@ function extractParagraphContent(paragraph, _documentCounts, emptyHyperlinks, te
 			content.push(createEmptyHyperlink(item.attrs));
 		}
 	};
+	/**
+	* Cut the emitted sequence where the wrapper the nodes sit inside changes.
+	*
+	* Nothing that spans nodes may span the cut: a run, a hyperlink and a
+	* revision each sit inside one wrapper or outside it, never half in. The
+	* comment ranges `syncCommentRanges` writes are cut with the rest, which
+	* puts a range that opens or closes inside a wrapper inside it, as Word
+	* writes it; a range marker says the same thing at either level, because the
+	* wrapper is transparent and the position is unchanged.
+	*/
+	const enterWrapperGroup = (node) => {
+		const stack = inlineWrapperStackOf(node);
+		const key = inlineWrapperStackKey(stack);
+		if (key === wrapperGroupKey) return;
+		flushCurrentInline();
+		currentTrackedChange = void 0;
+		wrapperGroupKey = key;
+		wrapperGroups.push({
+			start: content.length,
+			stack
+		});
+	};
 	const processInlineNode = (node, offset) => {
 		if (node.type.name === "renderedPageBreak" && leadingRenderedPageBreakPending) {
 			leadingRenderedPageBreakPending = false;
 			return;
 		}
 		leadingRenderedPageBreakPending = false;
+		enterWrapperGroup(node);
 		syncCommentRanges(node, offset);
+		if (node.type.name === "commentReference") {
+			flushCurrentInline();
+			currentTrackedChange = void 0;
+			content.push({
+				type: "commentReference",
+				id: expectCommentReferenceAttrs(node).commentId
+			});
+			return;
+		}
 		const linkMark = node.marks.find((m) => m.type.name === "hyperlink");
 		const noteRefMark = node.marks.find((m) => m.type.name === "footnoteRef");
 		const insertionMark = node.marks.find((m) => m.type.name === "insertion");
@@ -1210,6 +1334,10 @@ function extractParagraphContent(paragraph, _documentCounts, emptyHyperlinks, te
 					});
 					return;
 				}
+				if (isInlineLevelPreservedXml(node)) {
+					currentTrackedChange.hyperlink.children.push(createPreservedInline(node));
+					return;
+				}
 				const run = createTrackedChangeRun({
 					...formattingContext,
 					marks: otherMarks,
@@ -1256,6 +1384,10 @@ function extractParagraphContent(paragraph, _documentCounts, emptyHyperlinks, te
 				currentTrackedChange.wrapper.content.push(createMathFromNode(node));
 				return;
 			}
+			if (isInlineLevelPreservedXml(node)) {
+				currentTrackedChange.wrapper.content.push(createPreservedInline(node));
+				return;
+			}
 			const run = createTrackedChangeRun({
 				...formattingContext,
 				marks: otherMarks,
@@ -1325,6 +1457,9 @@ function extractParagraphContent(paragraph, _documentCounts, emptyHyperlinks, te
 		} else if (node.type.name === "symbol") {
 			flushCurrentInline();
 			content.push(createSymbolRun(node, node.marks, formattingContext));
+		} else if (node.type.name === "preservedXml") {
+			flushCurrentInline();
+			content.push(isInlineLevelPreservedXml(node) ? createPreservedInline(node) : createPreservedXmlRun(node, node.marks, formattingContext));
 		} else if (node.type.name === "hardBreak") {
 			flushCurrentInline();
 			content.push(createBreakRun(expectHardBreakAttrs(node), node.marks, formattingContext));
@@ -1379,11 +1514,11 @@ function extractParagraphContent(paragraph, _documentCounts, emptyHyperlinks, te
 	});
 	flushEmptyHyperlinksThroughOffset(Number.POSITIVE_INFINITY);
 	flushCurrentInline();
-	for (const commentId of openedComments) content.push({
+	for (const commentId of byCommentId(openedComments)) content.push({
 		type: "commentRangeEnd",
 		id: commentId
 	});
-	return content;
+	return nestInlineWrapperGroups(content, wrapperGroups);
 }
 function createTrackedChangeRun({ baseParagraphFormatting, inheritedFormatting, marks, node, paragraphMarkFormatting, paragraphMarkPrecedesStyle, styleResolver }) {
 	const formattingContext = {
@@ -1408,6 +1543,7 @@ function createTrackedChangeRun({ baseParagraphFormatting, inheritedFormatting,
 		};
 		restoreRunPropertyChanges(run, marks);
 	} else if (node.type.name === "symbol") run = createSymbolRun(node, marks, formattingContext);
+	else if (node.type.name === "preservedXml") run = createPreservedXmlRun(node, marks, formattingContext);
 	else if (node.type.name === "hardBreak") run = createBreakRun(expectHardBreakAttrs(node), marks, formattingContext);
 	else if (node.type.name === "pageBreakRun") {
 		const { clear } = expectPageBreakRunAttrs(node);
@@ -1582,6 +1718,16 @@ function addNodeToHyperlink({ baseParagraphFormatting, hyperlink, inheritedForma
 		}));
 		return;
 	}
+	if (node.type.name === "preservedXml") {
+		hyperlink.children.push(isInlineLevelPreservedXml(node) ? createPreservedInline(node) : createPreservedXmlRun(node, nonLinkMarks, {
+			baseParagraphFormatting,
+			inheritedFormatting,
+			paragraphMarkFormatting,
+			paragraphMarkPrecedesStyle,
+			styleResolver
+		}));
+		return;
+	}
 	if (node.type.name === "hardBreak") {
 		hyperlink.children.push(createBreakRun(expectHardBreakAttrs(node), nonLinkMarks, {
 			baseParagraphFormatting,
@@ -1679,6 +1825,44 @@ function createSymbolRun(node, marks, formattingContext) {
 	restoreRunPropertyChanges(run, marks);
 	return run;
 }
+/**
+* Whether the atom's markup is a paragraph child rather than a run child.
+*
+* The two are the same node in the editor and different elements in the file:
+* `w:ruby` has to go back inside a `w:r` and `w:permStart` may not, because
+* the schema admits no such child of a run and Word reports the package as
+* unreadable content.
+*/
+const isInlineLevelPreservedXml = (node) => node.type.name === "preservedXml" && expectPreservedXmlAttrs(node).level === PRESERVED_XML_LEVELS.inline;
+/** The paragraph-level capture an inline-level atom writes back. */
+const createPreservedInline = (node) => {
+	const { xml, text } = expectPreservedXmlAttrs(node);
+	return {
+		type: "preservedInline",
+		xml,
+		text
+	};
+};
+/**
+* Rebuild the run around a preserved child. The markup is opaque and comes
+* back byte for byte; only the run properties around it are rebuilt from the
+* atom's marks, exactly as for a symbol.
+*/
+function createPreservedXmlRun(node, marks, formattingContext) {
+	const { xml, text } = expectPreservedXmlAttrs(node);
+	const run = {
+		type: "run",
+		content: [{
+			type: "preservedXml",
+			xml,
+			text
+		}]
+	};
+	const formatting = getAtomRunFormattingFromMarks(marks, formattingContext);
+	if (formatting) run.formatting = formatting;
+	restoreRunPropertyChanges(run, marks);
+	return run;
+}
 function restoreRunPropertyChanges(run, marks) {
 	const changeMark = marks.find((mark) => mark.type.name === "runPropertyChange");
 	if (!changeMark) return;
@@ -1821,7 +2005,7 @@ function createFieldFromNode(node, { baseParagraphFormatting, inheritedFormattin
 		paragraphMarkFormatting,
 		paragraphMarkPrecedesStyle: paragraphMarkPrecedesStyle ?? false,
 		styleResolver: styleResolver ?? null
-	}).filter((content) => content.type === "run" || content.type === "hyperlink");
+	}).filter((content) => content.type === "run" || content.type === "hyperlink" || content.type === "preservedInline");
 	const displayText = attrs.displayText ?? "";
 	const displayRun = {
 		type: "run",
@@ -1842,8 +2026,8 @@ function createFieldFromNode(node, { baseParagraphFormatting, inheritedFormattin
 			fieldCode: [],
 			fieldResult: fieldContent.length > 0 ? fieldContent.filter((content) => content.type === "run") : fallbackFieldContent
 		};
-		if (attrs.fldLock) complex.fldLock = true;
-		if (attrs.dirty) complex.dirty = true;
+		if (attrs.fldLock !== void 0) complex.fldLock = attrs.fldLock;
+		if (attrs.dirty !== void 0) complex.dirty = attrs.dirty;
 		return complex;
 	}
 	const simple = {
@@ -1852,8 +2036,8 @@ function createFieldFromNode(node, { baseParagraphFormatting, inheritedFormattin
 		fieldType: attrs.fieldType,
 		content: fieldContent.length > 0 ? fieldContent : fallbackFieldContent
 	};
-	if (attrs.fldLock) simple.fldLock = true;
-	if (attrs.dirty) simple.dirty = true;
+	if (attrs.fldLock !== void 0) simple.fldLock = attrs.fldLock;
+	if (attrs.dirty !== void 0) simple.dirty = attrs.dirty;
 	return simple;
 }
 const synchronizeFieldDisplayText = (content, displayText, fallbackRun) => {
@@ -1864,6 +2048,7 @@ const synchronizeFieldDisplayText = (content, displayText, fallbackRun) => {
 				visit(child);
 				continue;
 			}
+			if (child.type !== "hyperlink") continue;
 			for (const hyperlinkChild of child.children) if (hyperlinkChild.type === "run") visit(hyperlinkChild);
 		}
 	};
@@ -1909,7 +2094,7 @@ function createInlineSdtFromNode(node, textBoxAnchorMarkers, formattingContext)
 	return {
 		type: "inlineSdt",
 		properties: sdtPropertiesFromAttrs(expectSdtAttrs(node)),
-		content: extractParagraphContent(node, void 0, void 0, textBoxAnchorMarkers, false, formattingContext).filter((c) => c.type === "run" || c.type === "hyperlink" || c.type === "simpleField" || c.type === "complexField" || c.type === "inlineSdt" || c.type === "insertion" || c.type === "deletion" || c.type === "moveFrom" || c.type === "moveTo" || c.type === "mathEquation")
+		content: extractParagraphContent(node, void 0, void 0, textBoxAnchorMarkers, false, formattingContext).filter(isInlineSdtContent)
 	};
 }
 /**
@@ -1942,9 +2127,13 @@ function createImageRun(node) {
 	if (attrs.docPrName != null) image.docPrName = attrs.docPrName;
 	if (attrs.alt != null) image.alt = attrs.alt;
 	if (attrs.title != null) image.title = attrs.title;
-	const imageTransform = parseTransformAttr(attrs.transform);
+	const imageTransform = readAuthoredTransform(attrs);
 	if (imageTransform) image.transform = imageTransform;
 	if (attrs.opacity != null) image.opacity = attrs.opacity;
+	if (attrs.brightness != null || attrs.contrast != null) image.effects = {
+		...attrs.brightness == null ? {} : { brightness: attrs.brightness },
+		...attrs.contrast == null ? {} : { contrast: attrs.contrast }
+	};
 	const imagePosition = imagePositionFromAttrs(attrs.position);
 	if (imagePosition) image.position = imagePosition;
 	if (attrs.layoutInCell !== void 0) image.layoutInCell = attrs.layoutInCell;
@@ -2049,7 +2238,7 @@ function createShapeRun(node) {
 	if (attrs.title !== void 0) shape.title = attrs.title;
 	const geometryAdjustments = parseShapeGeometryAdjustments(attrs.geometryAdjustments);
 	if (geometryAdjustments !== void 0) shape.geometryAdjustments = geometryAdjustments;
-	const shapeTransform = parseTransformAttr(attrs.transform);
+	const shapeTransform = readAuthoredTransform(attrs);
 	if (shapeTransform) shape.transform = shapeTransform;
 	const wrap = { type: attrs.wrapType || "inline" };
 	assignWrapDistances(wrap, attrs);
@@ -2860,6 +3049,7 @@ function convertPMTableRow(node, documentCounts, activeVerticalMerges, styleReso
 	const rowFormatting = tableRowAttrsToFormatting(attrs);
 	if (rowFormatting) row.formatting = rowFormatting;
 	if (Array.isArray(attrs.trPrChange) && attrs.trPrChange.length > 0) row.propertyChanges = [...attrs.trPrChange];
+	if (attrs._preservedAttributes && attrs._preservedAttributes.length > 0) row.preservedAttributes = attrs._preservedAttributes;
 	if (attrs.trIns) row.structuralChange = {
 		type: "tableRowInsertion",
 		info: {
@@ -2952,6 +3142,13 @@ function convertPMTableCell(node, documentCounts, styleResolver = null) {
 			textBoxAnchorMarkers,
 			styleResolver
 		});
+		else if (contentNode.type.name === "preservedBlock") {
+			content.push({
+				type: "preservedBlock",
+				xml: expectPreservedBlockAttrs(contentNode).xml
+			});
+			previousStandaloneTextBox = null;
+		}
 	});
 	removeUnresolvedTextBoxAnchors(content, textBoxAnchorMarkers);
 	const cell = {
@@ -3086,7 +3283,7 @@ function convertPMTextBox(node, styleResolver = null) {
 	if (attrs.textBoxName !== void 0) shape.name = attrs.textBoxName;
 	if (attrs.alt !== void 0) shape.alt = attrs.alt;
 	if (attrs.title !== void 0) shape.title = attrs.title;
-	const transform = parseTransformAttr(attrs.transform);
+	const transform = readAuthoredTransform(attrs);
 	if (transform) shape.transform = transform;
 	if (attrs.fillColor) shape.fill = {
 		type: "solid",
@@ -3156,8 +3353,10 @@ function updateDocumentContent(originalDocument, pmDoc) {
 * Used for converting edited header/footer PM content back to the document
 * model.
 */
-function proseDocToBlocks(pmDoc, baseContent, styles, options) {
-	const blocks = extractBlocks(pmDoc, "resolve", styles ? createStyleEngine(styles) : null, options?.emptyFieldResult ?? "serializerFallback");
+function proseDocToBlocks(pmDoc, baseContent, styles) {
+	const blocks = extractBlocks(pmDoc, "resolve", styles ? createStyleEngine(styles) : null);
+	joinCommentRangesAcrossParagraphs(blocks);
+	completeCommentReferences(blocks);
 	const linkedSources = restoreLinkedParagraphPropertySources(blocks);
 	if (baseContent) restoreParagraphPropertySources(blocks, baseContent, linkedSources.targets, linkedSources.sources);
 	return blocks;

package/dist/prosemirror/conversion/index.d.ts

@@ -1,4 +1,4 @@
-import { EmptyFieldResultMode, ProseDocToBlocksOptions, fromProseDoc, proseDocToBlocks, updateDocumentContent } from "./fromProseDoc.js";
+import { fromProseDoc, proseDocToBlocks, updateDocumentContent } from "./fromProseDoc.js";
 import { ToProseDocOptions, createEmptyDoc, headerFooterToProseDoc, toProseDoc } from "./toProseDoc.js";
 import { ProseMirrorDocumentValidationIssue, ValidateProseMirrorDocumentResult, assertValidProseMirrorDocument, formatProseMirrorDocumentIssues, validateProseMirrorDocument } from "../validation.js";
-export { type EmptyFieldResultMode, type ProseDocToBlocksOptions, type ProseMirrorDocumentValidationIssue, type ToProseDocOptions, type ValidateProseMirrorDocumentResult, assertValidProseMirrorDocument, createEmptyDoc, formatProseMirrorDocumentIssues, fromProseDoc, headerFooterToProseDoc, proseDocToBlocks, toProseDoc, updateDocumentContent, validateProseMirrorDocument };
+export { type ProseMirrorDocumentValidationIssue, type ToProseDocOptions, type ValidateProseMirrorDocumentResult, assertValidProseMirrorDocument, createEmptyDoc, formatProseMirrorDocumentIssues, fromProseDoc, headerFooterToProseDoc, proseDocToBlocks, toProseDoc, updateDocumentContent, validateProseMirrorDocument };

package/dist/prosemirror/conversion/toProseDoc.d.ts

@@ -1,16 +1,9 @@
 import { document_d_exports } from "../../types/document.js";
 import { StyleEngine } from "../../style-engine/styleEngine.js";
 import "../../style-engine/index.js";
+import { ParseContext } from "../../docx/parseContext.js";
 import { Node } from "prosemirror-model";
 //#region src/prosemirror/conversion/toProseDoc.d.ts
-type UnsupportedDocxToProseMirrorOwner = "complex-field-instruction" | "field-result" | "page-break-bearing-run" | "paragraph-borders" | "paragraph-frame" | "paragraph-outline" | "paragraph-text-box-anchor" | "table-cell" | "text-box";
-declare const UnsupportedDocxToProseMirrorConversionError_base: import("better-result").TaggedErrorClass<"UnsupportedDocxToProseMirrorConversionError">;
-/** DOCX content that cannot be preserved by the editable ProseMirror model. */
-declare class UnsupportedDocxToProseMirrorConversionError extends UnsupportedDocxToProseMirrorConversionError_base<{
-  message: string;
-  owner: UnsupportedDocxToProseMirrorOwner;
-  contentType: document_d_exports.RunContent["type"];
-}> {}
 /**
  * Options for document conversion
  */
@@ -19,6 +12,11 @@ type ToProseDocOptions = {
   styles?: document_d_exports.StyleDefinitions;
   /** Theme used when converting themed table/cell values in nested content. */
   theme?: document_d_exports.Theme | null;
+  /**
+   * Where a projection approximation is reported, on the channel the parsers
+   * already use. Absent means nobody is listening, not that nothing happened.
+   */
+  warn?: ParseContext["warn"];
 };
 /**
  * Convert a Document to a ProseMirror document
@@ -53,4 +51,4 @@ declare function footnoteToProseDoc(content: document_d_exports.BlockContent[],
  */
 declare function createEmptyDoc(): Node;
 //#endregion
-export { ToProseDocOptions, UnsupportedDocxToProseMirrorConversionError, createEmptyDoc, footnoteToProseDoc, headerFooterToProseDoc, resolveParagraphDefaultTextFormatting, standaloneTableCellToProseMirror, toProseDoc };
+export { ToProseDocOptions, createEmptyDoc, footnoteToProseDoc, headerFooterToProseDoc, resolveParagraphDefaultTextFormatting, standaloneTableCellToProseMirror, toProseDoc };